Datasheet for local AI49 models98 GPUsData read 2026-09-25
No adsNo tracking
Check my GPU →

Z-Image Turbo on 8 GB of VRAM

Image model · 6BAny 8 GB GPUData 2026-09-25
Runs
Yes, with a compressed file.

Download Q6_K (5.9 GB). With the model's working memory it needs about 7.9 GB, leaving 0.1 GB spare on 8 GB. Quality: very close to the original. For better quality, INT8 (6.2 GB) also runs, with about 0.2 GB spilling into system RAM — a little slower, still practical.

Best fileQ6_K
File size5.9 GB
VRAM needed~7.9 GB
System RAM32 GB+
Memory map · 8 GB GPU7.9 GB / 8 GB
04 GB8 GB
Weights Q6_K · 5.9 GBWorking memory · 1.2 GBReserve · 0.8 GBFree · 0.1 GB
calculated from real file sizes plus working memory. Not yet measured on this setup. How this works.

Worth trying on NVIDIA GPUs: with an up-to-date ComfyUI, Dynamic VRAM (on by default for NVIDIA since March 2026) streams whatever does not fit from system RAM, and ComfyUI's own start-up message recommends native FP8/INT8 files over GGUF, saying they “will be faster even if they are larger than your memory”. So before settling for a heavily compressed GGUF, try the native INT8 file (6.2 GB). The file recommended above is the best one that fits entirely — the safe choice on older ComfyUI versions, AMD and Intel. Comfy blog: Dynamic VRAM →

01What to download

PartFileFolderSize
Modelz-image-turbo-Q6_K.gguf
Q6_K · unsloth/Z-Image-Turbo-GGUF
models/unet5.9 GBDownload →
Text encoderqwen_3_4b_fp8_mixed.safetensors
Qwen3 4B FP8 · Comfy-Org/z_image_turbo
models/text_encoders5.6 GBDownload →
VAEae.safetensors
FLUX.1 VAE (ae) · Comfy-Org/z_image_turbo
models/vae0.3 GBDownload →
Total download · keep about the same free on disk11.9 GB

Alternatives: Qwen3 4B BF16 (8.0 GB). The 16-bit text encoder is a little more faithful; the smaller one is chosen here because it loads faster and needs less RAM. Sizes read from Hugging Face (2026-09-25). “Download” links start the file directly; the file name links are the same files the official ComfyUI workflows use.

System RAM: 32 GB or more. ComfyUI keeps the model file, the text encoder and the VAE in system RAM and moves them to the GPU as needed. With this set of files that is about 11.9 GB, plus roughly 6 GB for Windows, ComfyUI and a browser: 17.9 GB in total. With less RAM it still runs, but Windows starts swapping to disk and loading gets very slow. calculated

02Every file, on 8 GB

FileSizeNeededOn this cardQualityDownload
16-bit
SAFETENSORS · Comfy-Org
12.3 GB14.3 GB6.3 GB too bigthe original weightsHugging Face →
Q8_0
GGUF · unsloth
7.2 GB9.2 GBspills 1.2 GBpractically identical to the originalHugging Face →
INT8
SAFETENSORS · Comfy-Org
6.2 GB8.2 GBspills 0.2 GBpractically identical to the originalHugging Face →
Q6_K ←
GGUF · unsloth
5.9 GB7.9 GBfits · 0.1 GB sparevery close to the originalHugging Face →
Q5_K_M
GGUF · unsloth
5.6 GB7.6 GBfits · 0.4 GB spareclose; small differences in fine detailHugging Face →
Q4_K_M
GGUF · unsloth
5.0 GB7.0 GBfits · 1.0 GB sparegood; some loss in fine detail and textHugging Face →
Q3_K_M
GGUF · unsloth
4.2 GB6.2 GBfits · 1.8 GB sparenoticeable loss of detailHugging Face →
Q2_K
GGUF · unsloth
3.6 GB5.6 GBfits · 2.4 GB spareheavy loss; a last resortHugging Face →

Sizes from the Hugging Face file listing, read 2026-09-25. “Needed” = file + 1.2 GB working memory for this model + 0.8 GB kept free for the system.

03The text encoder

Qwen3 4B: 8.0 GB as 16-bit, 5.6 GB as FP8. ComfyUI encodes the prompt first and can push the encoder out of VRAM before sampling, so it does not have to fit together with the model. On 8 GB the FP8 encoder fits on its own, so prompt encoding stays fast.